[ppml] Summary of Trial Balloons for Dealing with IPv4AddressCountdown

[Stephen Sprunk]

Thus spake "Andy Davidson" <andy at nosignal.org>
> Do we have any empirical research of this, i.e. can someone
> who uses large chunks of v4 calculate a model that shows what
> the effect on the routing table would have been if they'd had to
> apply for smaller chunks of non-contiguous space more
> regularly?

Logically, if one multiplies the number of requests needed by N, then the 
number of required routing slots will multiply by N as well.  The only 
exception would be if allocations were sparse enough that subsequent 
requests would allow aggregation, but given the rate we're going through 
/8s, allocations can't be sparse enough for this to be true in enough cases 
to matter.

> Room to breathe in your v4 allocations could have also
> encouraged sloppiness and wastefulness.

Any org that has to come back for subsequent allocations shouldn't be 
wasteful or sloppy, because policy says they can't get any new space until 
they efficiently utilize their existing space.

> (Do we think all of - picking on some companies at random - all
> of 9/8, 13/8, 15/8, 44/8, etc.,etc. are effective use of large
> assignments..)

I don't think anyone doubts that excessive direct assignments are bad, 
particularly the legacy /8s.  In theory, the same effect as above applies, 
but in practice if one gives a slow-growing org more space than they'll ever 
need, they can't efficiently utilize it.  Their growth will lead to gradual 
efficiency improvements over time, but it'll never reach the level of an org 
that's given only what they need or that grows quickly, requiring multiple 
requests and thus audits.

I've worked with dozens of folks that used 10/8 internally, but most of them 
wouldn't have qualified for more than a /16 today if they had to follow 
ARIN's rules.  Many of those orgs are comparable in size to other orgs that 
adopted IPv4 early and got legacy /8 assignments.  Ditto for smaller folks 
that got legacy /16 assignments; many wouldn't qualify for more than a 
handful of /24s today.

> A big routing table is bad, we all agree.  Running out of v4 is
> really really bad.

Both are going to happen, no matter what we do.  Given that we're all going 
to have to pay to move to IPv6 sooner or later, how much justification is 
there for collectively spending billions of dollars making IPv4 limp along 
another few years?

S

Stephen Sprunk      "Those people who think they know everything
CCIE #3723         are a great annoyance to those of us who do."
K5SSS                                             --Isaac Asimov